Two-dimensional semi-parametric alignment of chromatograms

• Published in: Journal of Chromatography A Vol. 1345; pp. 193 - 199
• Main Authors: de Boer, Wim P.H., Lankelma, Jan
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 06.06.2014; Elsevier
• Subjects: Algorithms; Alignment; Animals; Applied sciences; breast neoplasms; Breast Neoplasms - chemistry; Chromatography, Gas - methods; Chromatography, Liquid - methods; Comprehensive; Crude oil, natural gas and petroleum products; data collection; Diesel; diesel fuel; Diesel fuels; Energy; Exact sciences and technology; Fuels; liquid chromatography; mass spectrometry; Mass Spectrometry - methods; Mathematical analysis; Mathematical models; Mice; Prospecting and production of crude oil, natural gas, oil shales and tar sands; Two dimensional; Warp; Warp function; Alignment; Comprehensive; Warp function; Two-dimensional; Biological fluid; Two dimensional chromatography; Healthy subject; Diesel fuel; Coupled method; Rodentia; HPLC chromatography; Patient; Chromatogram; Malignant tumor; Algorithm; Blood; Gas chromatography; Vertebrata; Mammalia; Mouse; Animal; Blood serum; Mass spectrometry; Cancer
• ISSN: 0021-9673; 1873-3778
• DOI: 10.1016/j.chroma.2014.04.034

•We have extended the semi-parametric alignment method to two-dimensions.•We have aligned challenging GC×GC diesel oil chromatograms with success.•The two-dimensional warp function is computed iteratively.•A computer program with user-friendly graphical interface is available. We present a comprehensive alignment algorithm that extends the semi-parametric approach to two dimensions. The algorithm is based on modeling shifts with a two-dimensional “warp function” such that the sample chromatogram – its shifts corrected with the warp function – is adjusted to the reference chromatogram by minimizing the squared intensity difference. A warp function approach has the advantage that overlapping peaks are easily dealt with compared to other proposed two-dimensional algorithms. Another advantage is that missing peaks are allowed if the absence of these peaks has little numerical effect on the warp function computation and if these peaks occur between existing peaks. Performance of the algorithm is demonstrated using GC×GC data from three batches of three diesel oil samples and LC-MS data from a mouse breast cancer data set.